Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

conv_cudnn_op not utilizing workspace #6320

Closed
tonyyang-svail opened this issue Dec 6, 2017 · 0 comments · Fixed by #6326
Closed

conv_cudnn_op not utilizing workspace #6320

tonyyang-svail opened this issue Dec 6, 2017 · 0 comments · Fixed by #6326
Assignees

Comments

@tonyyang-svail
Copy link

tonyyang-svail commented Dec 6, 2017

The default value of workspace_size_MB has been set to 4096.

AddAttr<int>("workspace_size_MB",
"workspace size for cudnn, in MB, "
"workspace is a section of GPU memory which will be "
"allocated/freed each time the operator runs, larger "
"workspace size can increase performance but also requires "
"better hardware. This size should be chosen carefully.")
.SetDefault(4096);

workspace_size_MB is of type int. It will overflow at

if (user_workspace_size > 0) {
workspace_size_limit = user_workspace_size * 1024 * 1024;
}

The similar problem appears at backward.

The test won't fail because workspace of size 0 is still a valid initialization. If we fix this we can get 50% performance improvements on ConvNet on TitanX. #6321

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging a pull request may close this issue.

2 participants